Performance evaluation of pansharpening for planetary exploration: A case study on the implementation of TGO CaSSIS with MRO HiRISE

Abstract

The present study analyses the potential of pansharpening algorithms for planetary exploration studies, testing their performance with the 4-band images from the Colour and Stereo Surface Imaging System (CaSSIS) aboard the Exomars 2016 Trace Grace Orbiter (TGO) using HiRISE images from the Mars Reconnaissance Orbiter (MRO) mission as the base. Due to the lack of suitable open-source tools, a suite of scripts was developed to improve alignment between images and enable different component substitution (CS) pansharpening algorithms. The tools developed were tested on a database of images encompassing several regions of Mars to explore its vast diversity in colours, tones, and textures. Then, the resulting images were investigated using spectral and structural performance indices, comparing the results with the source images and the colour information from the HiRISE central channels.

The results show that a substantial number of the tested algorithms are more than suitable for data enhancement, showing a considerable improvement in the structural characteristics of the images without sacrificing their spectral characteristics. In detail, the Gram-Schmidt method, widely used in terrestrial pansharpening, turns out to be the best compromise among the tested algorithms. Regarding the other tested algorithms, GIHS and the MMSE Brovey, a modified version of the classic Brovey, show the most significant increase in structural properties, while GHPF and GHPM show the interesting ability to maintain virtually unchanged spectral conditions of the multispectral source data.

In addition, the analysis reveals the applicability of pansharpening at a ground resolution increment up to 18 times, from 4.5 up to 0.25 m/px, a broader range than is usually used in traditional pansharpening.